VQV Signal
According to a post on X by @huang_biwei, AetherLab's CausalWAM adds an action expert to its CausalWM video model, so one model outputs future video frames and robot action chunks together. The model predicts motion first, then 3D geometry, then future frames. It is trained across robots using a shared end-effector action space on 15.8k curated hours of data, and the post says it reaches 93.0% average success on LIBERO-Plus after fine-tuning.
According to a post on X by @huang_biwei, AetherLab's CausalWAM adds an action expert to its CausalWM video model, so one model outputs future video frames and robot action chunks together. The model predicts motion first, then 3D geometry, then future frames. It is trained across robots using a shared end-effector action space on 15.8k curated hours of data, and the post says it reaches 93.0% average success on LIBERO-Plus after fine-tuning.
AI-assisted summary based on the listed source.
AetherLab's CausalWAM adds an action expert to its CausalWM video model, which predicts motion, then 3D geometry, then future frames, so one model outputs future frames and action chunks together. It trains across robots with a shared end-effector action space on 15.8k curated hours, and reaches 93.0% average...
Robotics teams and researchers following embodied AI would be most affected, since the approach links video prediction and robot control in a single model. The 93.0% result is reported on LIBERO-Plus, a benchmark, and the snippet does not describe real-world results. Independent evaluations and tests outside the benchmark are the next things to watch.
Hardware and robotics watchers may want to track whether this becomes a product, benchmark, or deployment signal.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to AI Video, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
